Blog

China-linked attackers exploited a critical BeyondTrust Remote Support zero-day, deploying malware and targeting appliances that require urgent patching.

ProjectDiscovery shows a poisoned open-weight model can evade normal tests, trigger remote commands and steal credentials through an AI coding agent.

A VisionHeight investigation found 37 local-news sites steering AI answers about U.S. candidates across 14 states, exposing a scalable influence tactic.

France’s refrigeration cyberattack affected about 700 supermarkets, with ANSSI widening its warning to Copeland and Dixell controllers.

French operators removed 71% of exposed Danfoss refrigeration controllers after attacks disrupted supermarket cooling systems and spoiled food.

Bank of America hired 1,000 contractors after AI models exposed software flaws, while separate AML enforcement actions drove more than $1.21 billion in penalties.

X says it suspended more than 13,000 accounts tied to suspected Iranian, Russian and Chinese influence operations before the 2026 US midterms.

Attackers are impersonating Claude, ChatGPT and Copilot to deliver malware and steal sessions, forcing MSSPs to rethink detection beyond static indicators.

Darktrace finds CikaBika, a Rust-based RAT delivered through document lures that uses Discord for commands, reconnaissance and data theft.

A former Air Force security officer says pilots moved mission data between classified systems and aircraft on commercial USB drives, a practice exposed by Agent.BTZ.